Naive Retries Cut Success From 55.4% to 41.5% Under Correlated Failures; 1 of 113 Production Configs Randomizes Delay

arXiv 2608.25403·medium signalXBlueskyLinkedInCopy link

Mehan and Saluja introduce the retry amplification factor and audit 200 open-source Python microservice projects. Explicit retry logic is detected in 11.5%, though their own false-negative audit puts true prevalence near 41%; among detected projects 60.9% have at least one configuration with no backoff, and exactly one of 113 production configurations randomizes its delay. In simulation (n=100 trials per strategy) a naive standard retry policy under correlated failure drops success from 55.4% to 41.5% versus not retrying at all, while their Adaptive Retry Budgeting holds near the no-retry baseline and still recovers transient faults.